Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
ex3ndr
on Nov 2, 2023
|
parent
|
context
|
favorite
| on:
M3 Max Geekbench
That's not bad actually because bottleneck in training is in RAM and we can have 128gb...
blovescoffee
on Nov 2, 2023
|
next
[–]
If you’re training a model that needs 128gb, the GPU speed is going to be a “bottleneck”
thangngoc89
on Nov 2, 2023
|
prev
[–]
I believe that Apple Silicone is still too slow for training. The big RAM is really useful for interference of big models though
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
